* implement restarting workers after certain number of task executions
* Clean up python code.
* Don't start new worker when an actor disconnects.
* Move wait_for_pid_to_exit to test_utils.py.
* Add test.
* Fix linting errors.
* Fix linting.
* Fix typo.
* add tf metrics
* comments
* fix network scopes
* add doc
* initial work
* try with 3 virtual cpus
* clean up metrics
* use format string
* fix trace level
* back to pong
* always run summary on cpu
* plot intermediate and final sgd stats
* add back a global step
* update
* add timeline
* use staging area and reuse weights properly
* stage at cpu
* whoops, stage only the batch
* clean up a bit
* fix py flake
* wip
* create an optimizer graph per device
* print timeline on 5th batch instead
* print examples per second
* log placement for training ops
* force placement on cpu:0
* try separating weights onto different gpus
* try using nccl
* add cpu fallback
* remove space from date
* check has gpu device
* fix flag config
* checkpoint
* wip
* update
* add some timing
* trace loading
* try cpu
* revert that
* remove expensive test
* lint
* cleanups
* clean up timers
* clean it up a bit
* fix code for non-scalar action spaces
* address some nits
* fix quotes
* efficient shuffling between sgd epochs
* local scheduler
* redirect output files to be associated with workers rather than the local scheduler
* fixed formatting
* fixes
* Moved output redirection logic to worker.py.
* Changed write mode.
* Fixed formatting.
* Added comment.
* Reuse log file creation in services.py.
* Fix linting.
* Fix problem in which multiple processes attempt to create /tmp/raylogs at the same time.
* Added to troubleshooting documentation about whether redefining remote functions runs the new code version
* Minor correction to troubleshooting documentation
* Writing new documentation page for using Ray with GPUs
* Wrote new documentation page on using ray with gpus
* Add some more details.
* Added to troubleshooting documentation about whether redefining remote functions runs the new code version
* Minor correction to troubleshooting documentation
* Small rewordings.
* Rewrote and reordered the examples in the Actor documentation for clarity. Also added an introduction to Gym
* Minor tweaks to actor documentation
* Small changes to wording.
* Fix bug in which plasma client file descriptors were not closed.
* Add logging statement when disconnecting client from plasma store.
* Fix after rebasing.
* Add more checks to plasma disconnect client.
* Add script for building MacOS wheels.
* Small cleanups to script.
* Fix setting of PATH before building wheel.
* Create symbolic link to correct Python executable so Ray installation finds the right Python.
* Address comments.
* Rename readme.
* Add manylinux setup
* Switch to cp27mu
* python/MANIFEST.in
* Fix MANIFEST.in
* Add build-wheel-manylinux1.sh
* Update readme
* Install correct version of numpy
* Fix typo in README-manylinux1.md
* Don't install cmake
* Remove commented line from setup.py
* Delete unused manylinux1.sh
* Run setup.py bdist_wheel twice
* Don't use package_data and MANIFEST.in.
* Small aesthetic change.
* Trigger build_ext in setup.py.
* Remove nonexistent file from MANIFEST.in.
* Manually copy files in MANIFEST.in to where Python expects them in order to prevent setup.py from having to be run twice.
* Only run setup.py once when building wheels.
* Aesthetic change to readme.
* Copy generated flatbuffer Python files in build_ext.
* Fix permission denied error by making sure to preserve executableness when copying files.
* Remove unnecessary argument to setup.py.
* Remove MANIFEST.in and move files to include into list in setup.py.
* Fix numpy version when building wheels and replace rm with git clean.
* Enable remote function and actor definitions to close over actor definitions.
* Give better error message if actor objects are pickled.
* Add tests for closing over actor definitions.
* Fix linting.
* add tf metrics
* comments
* fix network scopes
* add doc
* use format string
* fix trace level
* plot intermediate and final sgd stats
* add back a global step
* commit jekyll template
* Port blog post to markdown.
* Small changes.
* Improvements to layout and post.
* More improvements.
* Add computation graph figures to the blog post.
* Small changes.
* Update gitignore.
* Implement sharding in the Ray core
* Single node Python modifications to do sharding
* Do the sharding in redis.cc
* Pipe num_redis_shards through start_ray.py and worker.py.
* Use multiple redis shards in multinode tests.
* first steps for sharding ray.global_state
* Fix problem in multinode docker test.
* fix runtest.py
* fix some tests
* fix redis shard startup
* fix redis sharding
* fix
* fix bug introduced by the map-iterator being consumed
* fix sharding bug
* shard event table
* update number of Redis clients to be 64K
* Fix object table tests by flushing shards in between unit tests
* Fix local scheduler tests
* Documentation
* Register shard locations in the primary shard
* Add plasma unit tests back to build
* lint
* lint and fix build
* Fix
* Address Robert's comments
* Refactor start_ray_processes to start Redis shard
* lint
* Fix global scheduler python tests
* Fix redis module test
* Fix plasma test
* Fix component failure test
* Fix local scheduler test
* Fix runtest.py
* Fix global scheduler test for python3
* Fix task_table_test_and_update bug, from actor task table submission race
* Fix jenkins tests.
* Retry Redis shard connections
* Fix test cases
* Convert database clients to DBClient struct
* Fix race condition when subscribing to db client table
* Remove unused lines, add APITest for sharded Ray
* Fix
* Fix memory leak
* Suppress ReconstructionTests output
* Suppress output for APITestSharded
* Reissue task table add/update commands if initial command does not publish to any subscribers.
* fix
* Fix linting.
* fix tests
* fix linting
* fix python test
* fix linting
